18:28:57 <vbatts|work> #startmeeting autuminal cleaning, for crosbymichael
18:28:57 <collabot> Meeting started Mon Oct  5 18:28:57 2015 UTC.  The chair is vbatts|work. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:28:57 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
18:28:57 <collabot> The meeting name has been set to 'autuminal_cleaning__for_crosbymichael'
18:29:14 <vbatts|work> there's not an "i" in autumnal
18:29:16 <vbatts|work> w/e
18:29:33 <vbatts|work> https://github.com/opencontainers/specs/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-asc
18:29:33 <crosbymichael> Lets start from the bottom up
18:29:37 <mrunalp> Sounds good
18:29:46 <crosbymichael> PRs, not issues?
18:29:58 <vbatts|work> k
18:30:05 <vbatts|work> https://github.com/opencontainers/specs/pulls?q=is%3Aopen+sort%3Aupdated-asc+is%3Apr
18:30:16 <vbatts|work> #topic least recently updated PRs
18:30:20 <vbatts|work> https://github.com/opencontainers/specs/pulls?q=is%3Aopen+sort%3Aupdated-asc+is%3Apr
18:30:23 <mrunalp> https://github.com/opencontainers/specs/pull/101
18:31:21 <mrunalp> crosbymichael: Either is fine
18:31:32 <mrunalp> I think we may be able to close out some issues too
18:31:43 <crosbymichael> we can start with PRs, then move onto issues
18:31:47 <wking> #101 still needs edits to the Go before merging, but review on the Markdown seemed like something that should happen first
18:32:06 <mrunalp> crosbymichael: Sounds good
18:32:16 <crosbymichael> Ok, lets review #101
18:32:56 <mrunalp> yeah
18:33:01 <mrunalp> It needs rebase
18:33:10 <mrunalp> Looking at the content now
18:33:33 <crosbymichael> ya, lets mostlly review content
18:34:26 <crosbymichael> wking: why would Go code need changed?
18:34:31 <wking> Go comments
18:35:42 <wking> but if we want to leave the required-ness out of the Go comments, we could leave them alone
18:37:00 <wking> although if, e.g. uid is optional, we probably can't use a uint32 for it (and would need to use a pointer, or special-case a particular value, or whatever to mark "not set")
18:37:41 <crosbymichael> its not optional
18:38:04 <crosbymichael> and it' does not need a pointer because the zero value is valid
18:38:28 <wking> it is optional, because it makes no sense for a cgroup-only entry
18:38:43 <crosbymichael> then ur pr is probably wrong
18:38:45 <wking> like this one: https://github.com/opencontainers/specs/pull/101/files#diff-428eec4013a655816cdefafd5d3505f1R87
18:39:27 <wking> possibly ;).  My preferred approach was #99, which separated cgroup handling from mknod handling
18:39:41 <wking> #101 just documents the current overloaded approach
18:40:09 <wking> what would uid be used for for that cgroup-only entry (https://github.com/opencontainers/specs/pull/101/files#diff-428eec4013a655816cdefafd5d3505f1R87)?
18:43:47 <wking> anyhow, back to work for me ;).  Good luck with the review :)
18:44:18 <mrunalp> crosbymichael: The question is whether we should combine the two or it makes more sense to separate
18:44:35 <crosbymichael> maybe if we want to review creation and permissions together being an issue we couldn't merge and look into that
18:45:35 <mrunalp> yeah, maybe punt on this one
18:45:45 <mrunalp> lk4d4: can give his thoughts when he has time :)
18:46:00 <vbatts> k. i'm back
18:47:11 <crosbymichael> next one
18:47:21 <crosbymichael> https://github.com/opencontainers/specs/pull/126
18:47:30 <vbatts|work> #topic https://github.com/opencontainers/specs/pull/126
18:47:57 <crosbymichael> -1 the file names are what they are supposed to be in Go
18:48:00 <crosbymichael> with _ not -
18:48:02 <vbatts|work> this one looks outdated
18:48:23 <mrunalp> yeah
18:48:28 <mrunalp> should close for now
18:48:30 <crosbymichael> even if it was updated still -1
18:48:40 <crosbymichael> you name files in go with _ not -
18:48:42 <vbatts|work> also, likely to change with my "top down" effort
18:48:44 <crosbymichael> because builds would work
18:48:47 <crosbymichael> yes
18:48:49 <mrunalp> crosbymichael: yep
18:49:13 <vbatts|work> want me to comment?
18:49:20 <crosbymichael> vbatts|work: yes
18:49:28 <crosbymichael> comment and close
18:50:24 <mrunalp> next https://github.com/opencontainers/specs/pull/127/files
18:50:30 <crosbymichael> #topic https://github.com/opencontainers/specs/pull/144
18:50:35 <crosbymichael> this one just needs a rebase
18:51:19 <mrunalp> yep
18:51:24 <mrunalp> What about 127?
18:52:40 <crosbymichael> maybe we should just make the change ourself for 127 and merge it and close
18:52:50 <mrunalp> crosbymichael: okay
18:53:35 <crosbymichael> i can do it
18:53:42 <mrunalp> alright
18:54:10 <crosbymichael> next
18:54:29 <mrunalp> #topic https://github.com/opencontainers/specs/pull/133/files
18:54:49 <mrunalp> I think this one probably needs more discussion on the mailing list.
18:54:58 <crosbymichael> i think that is being discussed on the mailing list and i don't see it being merge yet
18:55:09 <mrunalp> okay next
18:55:13 <crosbymichael> so close
18:55:14 <crosbymichael> ?
18:55:20 <vbatts|work> that's tough.
18:55:41 <crosbymichael> what is?
18:55:59 <vbatts|work> there is a story there. but this mounting of qcow right in the config seems too much
18:56:05 <mrunalp> yes
18:56:13 <mrunalp> can open a new one once there is agreement
18:56:35 <vbatts|work> perhaps there is a space for vendor custom config, but otherwise there the standard minimal core
18:57:12 <crosbymichael> eww
18:57:22 <mrunalp> Nothing blocks someone from encoding the source path today. Just the language and long direction need to be figured out w.r.t. hypervisors.
18:58:54 <mrunalp> closing/ not closing?
18:59:19 <mrunalp> I am okay keeping it as is for now. We can revisit it next time we do a sweep.
19:00:16 <crosbymichael> it would be nice to keep the issue tracker and PRs under control
19:00:29 <mrunalp> okay, let's close it then
19:00:33 <mrunalp> I can comment and close
19:00:34 <crosbymichael> makes it easier for reviewers and new ppl
19:00:36 <vbatts|work> i did
19:00:40 <vbatts|work> next
19:00:41 <mrunalp> :D
19:00:57 <crosbymichael> we don't want the type of volume that docker has, 150 open PRs and 75-120 new / week
19:01:27 <mrunalp> Makes sense
19:01:59 <crosbymichael> #topic https://github.com/opencontainers/specs/pull/142
19:02:01 <crosbymichael> LGTM
19:02:37 <mrunalp> +1
19:03:13 <vbatts|work> next
19:03:24 <vbatts|work> #topic https://github.com/opencontainers/specs/pull/171/files
19:04:01 <crosbymichael> up to u vbatts|work
19:04:16 <mrunalp> Looks like it should be split up and updated
19:04:19 <vbatts|work> hmm. the rebase needed makes this confusing
19:04:23 <mrunalp> Seems all over the place
19:05:10 * vbatts|work compares to master
19:06:38 <vbatts|work> but the namespace mapping isn't in the runtime. it is in the config
19:06:52 <vbatts|work> ... but _should_ that be runtime specific?
19:07:17 <mrunalp> yes
19:07:20 <vbatts|work> perhaps the container-config makes the ask for a uid/gid to be namespaced
19:07:32 <vbatts|work> and the runtime-config records the uid/gid it is mapped too?
19:08:04 <crosbymichael> i g2g
19:08:06 <mrunalp> vbatts|work: No, the mappings are exact
19:08:13 <mrunalp> crosbymichael: Alright cyl
19:08:16 <crosbymichael> whatever you two decide will be perfect
19:08:17 <crosbymichael> ;)
19:08:20 <mrunalp> Ha :D
19:08:20 <vbatts|work> heh
19:08:41 <mrunalp> vbatts|work: It isn't like cgroups..
19:08:47 <vbatts|work> mrunalp: so if 0 in the container is 2000 on the host, it will have to be 2000 on every/any host?
19:09:10 <mrunalp> yes, according to our config
19:09:12 <vbatts|work> its seems like it could eventually be a relative configuration.
19:09:36 <mrunalp> Still a runtime decision I think
19:10:17 <vbatts|work> but then the immutable config ought not choose the uid/gid on the host which would be mapped to
19:11:33 <vbatts|work> i could imagine some priv-escalation, where a container could specifiy that it is to be mapped to ... uid 62, and then if somehow escalated it could access anything httpd has rights to
19:11:37 <vbatts|work> who knows
19:12:02 <vbatts|work> mrunalp: want to comment on this PR?
19:12:28 <mrunalp> vbatts|work: Sure
19:12:58 <mrunalp> vbatts|work: I have a call coming up but I will comment on this PR. Maybe have another session later or tomorrow?
19:13:21 <vbatts|work> k
19:13:25 <vbatts|work> #endmeeting